Low accuracy of Bayesian latent class analysis for estimation of herd-level true prevalence under certain disease characteristics-An analysis using simulated data

Estimation of diagnostic sensitivity and specificity of abattoir registrations and bulk tank milk ELISA as herd-level tests for Fasciola hepatica using Bayesian latent class modelling

The common liver fluke, Fasciola hepatica, is a trematode parasite found worldwide, typically with a focal distribution due to its requirement for suitable climatic and environmental conditions to complete its lifecycle. Bovine fasciolosis causes suboptimal production and economic losses, including liver condemnation at slaughter. The lack of reliable diagnostic methods is a disadvantage to the increasing demand for surveillance and control. The aim of this study was to evaluate the diagnostic accuracy of bulk tank milk (BTM) antibody testing and aggregated abattoir registrations (AAR) of liver fluke as herd-level tests for F. hepatica infection using Bayesian latent class models. Data from the abattoirs in 2019-2021 and BTM, sampled in the winter of 2020/2021, were collected from 437 herds on the southwest coast of Norway. The BTM samples were analysed with the SVANOVIR® F. hepatica-Ab ELISA test, with results given as an optical density ratio (ODR), and later dichotomized using the recommended cut-off value from the test manufacturer (ODR ≥0.3). Based on the BTM ELISA test, 47.8% of the herds tested positive. The AAR test was defined as the herd-level proportion of female slaughtered animals registered with liver fluke infection during the study period. For this test, three cut-offs were used (a proportion of 0.05, 0.1 and 0.2). The herds were split into two subpopulations ("Coastal" and "Inland"), which were expected to differ in true prevalence of F. hepatica infection based on climate-related and geographical factors. The diagnostic accuracies of both tests were estimated using Bayesian latent class models with minimally informative priors. Post-hoc analysis revealed that the maximum sum of sensitivity (Se) and specificity (Sp) of the tests was achieved with a herd-level proportion of ≥0.1 registered with liver fluke as the AAR test. Using this cut-off, the median estimate for the diagnostic accuracy of the BTM ELISA was 90.4% (84.0-96.2 95% Posterior Credible Interval (PCI)) for Se and 95.3% (90.6-100% PCI) for Sp, while the median estimate of Se for AAR was 87.5% (81.4-93.1% PCI) and the median estimate of Sp for AAR was 91.0% (85.2-96.5% PCI). The cut-off evaluation of the SVANOVIR® F. hepatica-Ab ELISA test for BTM confirmed the manufacturer's recommended cut-off of ODR ≥0.3 to denote positive and negative herds. This study suggests that AAR and BTM ELISA test can be used as herd-level tools to monitor liver fluke infection, so that appropriate interventions against infection can be implemented as necessary.

Diagnostic performance of Mycoplasmopsis bovis antibody ELISA tests on bulk tank milk from dairy herds

BACKGROUND

Testing of bulk tank milk (BTM) for Mycoplasmopsis bovis (previously Mycoplasma bovis) antibodies is increasingly popular. However the performance of some commercially available tests is unknown, and cutoff values possibly need to be adjusted in light of the purpose. Therefore, the aim of this study was to compare the diagnostic performance of three commercially available M. bovis antibody ELISAs on BTM, and to explore optimal cutoff values for screening purposes. A prospective diagnostic test accuracy study was performed on 156 BTM samples from Belgian and Swiss dairy farms using Bayesian Latent Class Analysis. Samples were initially classified using manufacturer cutoff values, followed by generated values.

RESULTS

Following the manufacturer's guidelines, sensitivity of 91.4%, 25.6%, 69.2%, and specificity of 67.2%, 96.8%, 85.8% were observed for ID-screen, Bio K432, and Bio K302, respectively. Optimization of cutoffs resulted in a sensitivity of 89.0%, 82.0%, and 85.5%, and a specificity of 83.4%, 75.1%, 77.2%, respectively.

CONCLUSIONS

The ID-screen showed the highest diagnostic performance after optimization of cutoff values, and could be useful for screening. Both Bio-X tests may be of value for diagnostic or confirmation purposes due to their high specificity.

Bayesian latent class modelling of true prevalence in animal subgroups with application to bovine paratuberculosis infection

The prevalence of an infectious disease of animals living in separate groups (e.g. herds) is naturally analyzed using a Bayesian hierarchical latent class model. We propose an extension to this methodology by including subgroup level prevalence measures within the groups of animals. As an application illustrating the merits of our methodology, we reassessed the prevalence of bovine paratuberculosis (PTBC) infection in Hungarian commercial dairy farms. Our aim was to consolidate previous findings using a large amount of recent data and priors based on historical data. To model the subgroup level infection prevalence within animal groups, we considered correlated prevalences following beta distributions derived from independent normally distributed random herd effects. In the application, infection status of herds was handled as latent classes, multiparous and primiparous cows as within-herd subgroups. The novel methodology allows us to estimate both the mean and median conditional within-herd true prevalence (CWHP) related to each animal subgroup as well as other measures characterizing the interrelation of subgroups. The results of the application aligned with the findings of the former PTBC study, while the more recent and considerably larger dataset and the use of historical priors increased the reliability of the results. The STAN and JAGS codes of the application are available in Supplementary material.

Bayesian diagnostic test evaluation and true prevalence estimation of mycoplasma bovis in dairy herds

The true prevalence of dairy cattle herds with M. bovis infections in the Netherlands is unknown. Previous attempts to estimate prevalences were hampered by the absence of a diagnostic serological test that was validated under field conditions. This study estimated sensitivity and specificity of two commercial serum ELISAs and the true M. bovis herd prevalence using different Bayesian latent class models. A total of 7305 serum samples from 415 randomly chosen dairy herds were collected in fall/winter 2019 and investigated for presence of antibodies against M. bovis using the BIO-K-260 ELISA from Bio-X. Serum samples from 100 of these herds were also tested with a second ELISA, from IDvet. A Bayesian latent class model using the paired test results estimated a sensitivity of 14.1% (95% Bayesian probability interval (BPI): 11.6-16.7%) for the Bio-X ELISA and a specificity of 97.2% (95% BPI: 95.9-98.4%). Sensitivity and specificity for the IDvet ELISA were estimated at 92.5% (95% BPI: 88.3-96.5%) and 99.3% (95% BPI: 98.7-99.8%), respectively. Also, Bio-X ELISA sensitivity was considerably higher with data from calves only and with data from a selection of herds with a clinical outbreak, whereas the IDvet ELISA sensitivity was fairly constant under these conditions. These differences in test sensitivity is expected to be related to an effect of time since infection. A second Bayesian model, applied on test results of all 415 herds, estimated a true herd prevalence of 69.9% (95% BPI: 62.7-77.6%), suggesting M. bovis in endemic amongst dairy cattle herds in the Netherlands. To what extent seropositive herds have experienced a clinical outbreak needs further investigation.

Prevalence of respiratory disease in Irish preweaned dairy calves using hierarchical Bayesian latent class analysis

Introduction

Bovine respiratory disease (BRD) has a significant impact on the health and welfare of dairy calves. It can result in increased antimicrobial usage, decreased growth rate and reduced future productivity. There is no gold standard antemortem diagnostic test for BRD in calves and no estimates of the prevalence of respiratory disease in seasonal calving dairy herds.

Methods

To estimate BRD prevalence in seasonal calving dairy herds in Ireland, 40 dairy farms were recruited and each farm was visited once during one of two calving seasons (spring 2020 & spring 2021). At that visit the prevalence of BRD in 20 calves between 4 and 6 weeks of age was determined using thoracic ultrasound score (≥3) and the Wisconsin respiratory scoring system (≥5). Hierarchical Bayesian latent class analysis was used to estimate the calf-level true prevalence of BRD, and the within-herd prevalence distribution, accounting for the imperfect nature of both diagnostic tests.

Results

In total, 787 calves were examined, of which 58 (7.4%) had BRD as defined by a Wisconsin respiratory score ≥5 only, 37 (4.7%) had BRD as defined by a thoracic ultrasound score of ≥3 only and 14 (1.8%) calves had BRD based on both thoracic ultrasound and clinical scoring. The primary model assumed both tests were independent and used informed priors for test characteristics. Using this model the true prevalence of BRD was estimated as 4%, 95% Bayesian credible interval (BCI) (1%, 8%). This prevalence estimate is lower or similar to those found in other dairy production systems. Median within herd prevalence varied from 0 to 22%. The prevalence estimate was not sensitive to whether the model was constructed with the tests considered conditionally dependent or independent. When the case definition for thoracic ultrasound was changed to a score ≥2, the prevalence estimate increased to 15% (95% BCI: 6%, 27%).

Discussion

The prevalence of calf respiratory disease, however defined, was low, but highly variable, in these seasonal calving dairy herds.

Estimation of the true prevalence of inaccurate artificial inseminations in Irish milk recording dairy cows using a Bayesian latent class analysis

Inaccurate artificial insemination (IAI) refers to an artificial insemination (AI) that is performed when a cow is not in oestrus. IAIs have economic impacts on the dairy industry through of semen wastage or iatrogenic pregnancy loss. However, few studies have quantified the prevalence of IAIs in a population. The primary objective of this prospective study was to estimate the cow-level true prevalence of IAIs in Irish milk recording dairy herds using a latent class model with a Bayesian framework. Milk samples were collected at a milk recording laboratory from 576 dairy cows in 125 herds who had received an AI on the same day they were sampled for routine milk constituent analysis. Milk progesterone (MP4) analysis was conducted on these samples using radioimmunoassay to determine the progesterone concentration. Fertility data (i.e., subsequent calving date) was retrospectively obtained from the Irish National Cattle Breeding Federation for milk sampled cows and an apparent conception (AC) to the sample AI was determined based on an estimated gestational range of 270-290 days. Both tests (MP4 and AC) were used in a latent class model to estimate the true prevalence of IAI. For the MP4 test, a concentration of ≥ 5 ng/mL in whole milk was deemed to be test positive while for the AC test, a cow that did not conceive to the sampled AI was deemed test positive. Prior information for prevalence of IAI was obtained from a literature review while MP4 sensitivity (Se) and specificity (Sp) were obtained from expert opinion. Non-informative priors were used for the Se and Sp of the AC test. Posterior inferences (median and 95 % Bayesian probability intervals; BPI) were obtained using the 'rjags' package in the R statistical software. In the final model, median cow-level true prevalence of IAI was 4.4 % (BPI; 1.7-9.0 %). Median Se and Sp estimates for MP4, were 83.0 % (BPI; 65.0-96.2 %) and were 97.4 % (BPI; 94.6-99.6 %), respectively. Median Se and Sp estimates for AC, were 64.8 % (BPI; 44.5-88.6 %) and 49.8 % (BPI; 45.3-54.1 %), respectively. The present study estimates that the overall cow-level true prevalence of IAI in Irish dairy cows is relatively low. This is the first study to report the cow-level true prevalence of IAI using a Bayesian latent class model.

Evaluation using latent class models of the diagnostic performances of three ELISA tests commercialized for the serological diagnosis of Coxiella burnetii infection in domestic ruminants

ELISA methods are the diagnostic tools recommended for the serological diagnosis of Coxiella burnetii infection in ruminants but their respective diagnostic performances are difficult to assess because of the absence of a gold standard. This study focused on three commercial ELISA tests with the following objectives (1) assess their sensitivity and specificity in sheep, goats and cattle, (2) assess the between- and within-herd seroprevalence distribution in these species, accounting for diagnostic errors, and (3) estimate optimal sample sizes considering sensitivity and specificity at herd level. We comparatively tested 1413 cattle, 1474 goat and 1432 sheep serum samples collected in France. We analyzed the cross-classified test results with a hierarchical zero-inflated beta-binomial latent class model considering each herd as a population and conditional dependence as a fixed effect. Potential biases and coverage probabilities of the model were assessed by simulation. Conditional dependence for truly seropositive animals was high in all species for two of the three ELISA methods. Specificity estimates were high, ranging from 94.8% [92.1; 97.8] to 99.2% [98.5; 99.7], whereas sensitivity estimates were generally low, ranging from 39.3 [30.7; 47.0] to 90.5% [83.3; 93.8]. Between- and within-herd seroprevalence estimates varied greatly among geographic areas and herds. Overall, goats showed higher within-herd seroprevalence levels than sheep and cattle. The optimal sample size maximizing both herd sensitivity and herd specificity varied from 3 to at least 20 animals depending on the test and ruminant species. This study provides better interpretation of three widely used commercial ELISA tests and will make it possible to optimize their implementation in future studies. The methodology developed may likewise be applied to other human or animal diseases.

An Iterative, Frequentist Approach for Latent Class Analysis to Evaluate Conditionally Dependent Diagnostic Tests

Latent class analysis is a well-established method in human and veterinary medicine for evaluating the accuracy of diagnostic tests without a gold standard. An important assumption of this procedure is the conditional independence of the tests. If tests with the same biological principle are used, this assumption is no longer met. Therefore, the model has to be adapted so that the dependencies between the tests can be considered. Our approach extends the traditional latent class model with a term for the conditional dependency of the tests. This extension increases the number of parameters to be estimated and leads to negative degrees of freedom of the model, meaning that not enough information is contained in the existing data to obtain a unique estimate. As a result, there is no clear solution. Hence, an iterative algorithm was developed to keep the number of parameters to be estimated small. Given adequate starting values, our approach first estimates the conditional dependencies and then regards the resulting values as fixed to recalculate the test accuracies and the prevalence with the same method used for independent tests. Subsequently, the new values of the test accuracy and prevalence are used to recalculate the terms for the conditional dependencies. These two steps are repeated until the model converges. We simulated five application scenarios based on diagnostic tests used in veterinary medicine. The results suggest that our method and the Bayesian approach produce similar precise results. However, while the presented approach is able to calculate more accurate results than the Bayesian approach if the test accuracies are initially misjudged, the estimates of the Bayesian method are more precise when incorrect dependencies are assumed. This finding shows that our approach is a useful addition to the existing Bayesian methods, while it has the advantage of allowing simpler and more objective estimations.

Stakeholder perceptions of non-regulatory bovine health issues in Ireland: past and future perspectives

Background

In recent years, there have been multiple (political, environmental, cultural) drivers of change in Irish agriculture, including the establishment of Animal Health Ireland (AHI) in 2009, to provide leadership of non-regulatory livestock health issues (diseases and conditions of livestock that are endemic in Ireland but which are not currently subject to international legislation). In this study, we describe the opinion of stakeholders (farmers, veterinary practitioners and agricultural industry professional service providers), elicited by means of a survey, on their perceptions of changes in selected non-regulatory bovine health issues over the last 10 years and priority issues relevant to non-regulatory bovine health to be tackled over the next 10 years.

Results

A total of 673 individuals participated in the online questionnaire. For the majority of the non-regulatory bovine health issues, most participants felt there had been improvements over the last 10 years. However, professional service providers were generally more conservative in their response to improvements on-farm compared to farmers. Several issues, particularly BVD and udder health/milk quality, were viewed more positively by all relevant respondents. There was reasonable agreement between responses from different respondent types and sectors regarding the top three priorities relevant to non-regulatory bovine animal health for the next 10 years in Ireland, which included antimicrobial resistance (highlighting measures to reduce both on-farm usage and resistance), anthelmintic resistance, greenhouse emissions and calf welfare.

Conclusions

The results are encouraging, demonstrating a perception of improvement in a number of non-regulatory bovine health issues in Ireland over the last ten years. With respect to the next 10 years, stakeholders prioritised antimicrobial and anthelmintic resistance, greenhouse gas emissions and calf welfare, which aligns closely with broader societal concerns. This information is useful to AHI, particularly with respect to future priorities. However, these concerns are broad in scope and will require further considerations, including collaborations, between AHI and partnering organisations. Given that there were differences between farmers and professional service providers in responses, it is useful to consider how the aims and the benefits of future AHI programmes are framed and communicated to all stakeholders.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13620-020-00178-8.

Estimation of the sensitivity and specificity of four serum ELISA and one fecal PCR for diagnosis of paratuberculosis in adult dairy cattle in New Zealand using Bayesian latent class analysis

In New Zealand, a new diagnostic approach for the control of paratuberculosis in mixed aged milking cows has been developed using a combination of ELISA and quantitative fecal PCR (f-qPCR). Our analysis was designed to evaluate performance of these individual tests in infected or infectious mixed aged cows across the prevalence of infection typically encountered on NZ dairy farms and calculate test accuracy when used as a screening test of serological ELISAs for four separate antigens read in parallel followed by a confirmatory quantitative f-qPCR test. Data from a cross-sectional study of 20 moderate prevalence herds was combined with existing data from 2 low and 20 high prevalence herds forming a dataset of 3845 paired serum and fecal samples. Incidence of clinical Johne's disease (JD) was used to classify herds into three prevalence categories. High (≥ 3% annual clinical JD for the last three years), moderate (<3 - 1%) and low (<1% incidence for at least the last five years). Positive tests were declared if> 50 ELISA units and f-qPCR at two cut-points (≥1 × 10⁴ genomes/mL or >1 × 10³ genomes/mL). Fixed Bayesian latent class models at both f-qPCR cut-points, accounted for conditional independence and paired conditional dependence. Mixed models at both f-qPCR cut-points, using a different mechanism to account for conditional dependencies between tests were also implemented. Models (24 in number) were constructed using OpenBUGS. The aim was to identify Mycobacterium avium subsp. paratuberculosis (MAP) infected cows that met at least one of two criteria: shedding sufficient MAP in feces to be detected by f-qPCR or mounting a detectable MAP antibody response. The best fit to the data was obtained by modelling pairwise dependencies between tests in a fixed model or by accounting for dependencies in a mixed model at a fecal cut-off of ≥1 × 10⁴ genomes/mL. Test performance differed with prevalence, but models were robust to prior assumptions. For the fixed model, at a prevalence of 0.29 (95 % probability interval (PI) = 0.25-0.33), as a screening plus confirmatory f-qPCR, post-test probability for disease in a positive animal was 0.84 (95 %PI = 0.80-0.88) and 0.16 (95 %PI = 0.15-0.18) for disease in a test negative animal. In low prevalence herds (0.01(95 %PI = 0.00-0.04)) the equivalent figures were 0.84 (95 %PI = 0.08-0.92) and 0.00 (95 %PI = 0.00-0.02). These results suggest this is a useful tool to control JD on dairy farms, particularly in herds with higher levels of infection, where the sampling and testing cost per animal is defrayed across more detected animals.

Bayesian estimation of the true prevalence of paratuberculosis in Hungarian dairy cattle herds

Paratuberculosis is a chronic incurable disease caused by Mycobacterium avium subsp. paratuberculosis (MAP), which leads to extensive economic losses on dairy farms, and may also pose serious public health risk to the consumers. The aim of our study was to estimate the true prevalence of paratuberculosis in commercial dairy cattle herds participating in a voluntary MAP testing programme that started in February 2018 in Hungary. Milk samples collected during official milk recording were used for MAP ELISA testing. A Bayesian two-stage hierarchical (herd and animal level) model was fitted to the data. Altogether, 26,437 cows from 51 herds were sampled, which represents 14.4 % of the Hungarian dairy cow population. The median herd size was 477 cows (interquartile range: 331-709). Each studied farm had at least one ELISA positive cow, resulting in a herd-level apparent prevalence of 100 %. The overall within herd apparent prevalence was 5.5 %. Herd-level true prevalence was estimated at 89.1 % [95 % credible interval (CrI): 80.3-95.6%]. Within the infected herds, the median animal-level true prevalence was 4.4 % (3.2-5.8%) for primiparous and 10.3 % (7.9-12.9%) for multiparous cows, respectively. The probability of having an animal-level true prevalence of at least 5% among primiparous cows, within infected herds, was 17.8 %. Similarly, the probability of having an animal-level true prevalence of at least 5% or 10 % among multiparous cows was 100 % and 56 %, respectively. Simulations assuming herd-level true prevalence varying from 50 to 100 % revealed high accuracy of our Bayesian model. Our study showed that a large percentage of the studied Hungarian dairy cattle herds was infected with MAP.

A review of paratuberculosis in dairy herds - Part 1: Epidemiology

Bovine paratuberculosis is a chronic infectious disease of cattle caused by Mycobacterium avium subspecies paratuberculosis (MAP). This is the first in a two-part review of the epidemiology and control of paratuberculosis in dairy herds. Paratuberculosis was originally described in 1895 and is now considered endemic among farmed cattle worldwide. MAP has been isolated from a wide range of non-ruminant wildlife as well as humans and non-human primates. In dairy herds, MAP is assumed to be introduced predominantly through the purchase of infected stock with additional factors modulating the risk of persistence or fade-out once an infected animal is introduced. Faecal shedding may vary widely between individuals and recent modelling work has shed some light on the role of super-shedding animals in the transmission of MAP within herds. Recent experimental work has revisited many of the assumptions around age susceptibility, faecal shedding in calves and calf-to-calf transmission. Further efforts to elucidate the relative contributions of different transmission routes to the dissemination of infection in endemic herds will aid in the prioritisation of efforts for control on farm.